As an engineering leader managing multiple projects, a big part of my job was keeping teams aligned on what we were building and why. We’d write it down, review it, and try to stay disciplined. But in practice, the handoff from product to engineering was messy. Product would do their best to get something "shovel-ready," engineering would start pulling on threads, and it would take weeks of back-and-forth before we could commit.
Specs drifted, decisions lived in Slack, and the "why" behind things ended up scattered across a dozen places.
Often, a year later, we’d come back to the same area of the product and realize how much context was gone. Docs were out of date, decisions were buried in old threads, and some details were never written down at all. So we’d rebuild the context, re-litigate old discussions, and rely on what people remembered (if they were even still at the company).
When AI tools started getting good (I first felt it when I began experimenting with GPT-3.5), I had a simple hope: help turning half-formed notes and back-and-forth into something I could share with a team and build from. But chat alone didn’t solve the real problem. I needed to make the same change across related docs, keep decisions tied to their rationale, and see exactly what changed so I could trust it.
Then coding agents started getting better and better. Code could move faster, which meant we hit the "what exactly are we building" questions sooner and more often. Product was already spread thin, and now there was even less slack for the constant clarification loop.
That’s what inspired the creation of Product Graph.
Product Graph is built for AI-first teams that need to move fast and stay aligned. It captures the what and why in a structured way, so requirements become durable, machine-readable, and usable by both humans and agents.
The AI assistant is part of that workflow, not bolted on afterward. You can start from a brain dump, a meeting transcript, or scattered notes and work toward clear, reviewable requirements. The assistant helps you fill gaps, keep language consistent, and update related docs without losing track of what changed.
We didn’t want “AI in a doc.” We wanted a workspace where context stays connected, changes are reviewable, and the output is something a team or a coding agent can pick up and execute without guessing.
